The ability to build a detailed model of an individual decision maker and to have that model interact with models of many other individual decision makers opens large areas of research in both the social sciences and in computational science. This dissertation is a successful exercise in building a quite sophisticated agent as a model of a self-interested decision maker, then having multiple instances of that model form useful organizations. The work reported on here is remarkable in that it carefully constructs agents as instances of self-interested entities, with all calculations in the models being primitive recursive with respect to natural resources, technologies, and operations the agents can perform. These agents exhibit self-control and self-management within a framework of limited rationality.
This work also produced a principled approach to building agents that form organizations only when the agents participating in the organization believe it is in their own computationally grounded self-interest to do so. Other researchers generally assign a role to each agent, even when such agents are called ``self-interested.''
Construction of a self-interested decision-making agent requires the integration of a large number of disparate elements into a single operating entity. These elements include activities drawn from the calculations of orthodox economic theories and from operations research, supplemented by data gathering activities associated with artificially intelligent agents. The integration required the refinement of existing data structures to solve novel problems (such as the use of heaps in the solution of a modification to the classic knapsack problem), the reduction of problem size to enable its timely and rapid solution by resource-bounded agents, and and the invention of new interpretations of previously known economic optimization techniques (such as how a bounded agent treats the inadvertent accumulations of assets).